‘When everyone’s responsible, no one is’: the AI champion problem

Find a champion. Protect 20% of their time. Wait for it to spread. One of the people in the room with me had run that play before, in the era of robotic process automation. Their leadership agreed to protect 20% of an AI champion’s time. A month later, the manager had taken it back. There was more work to do.

That story came from an operations leader at Zapier Outpost Windsor, an invite-only retreat the automation company ran at Oakley Court, on the Thames near Windsor, from 8 to 10 September. Zapier covered flights, lodging and meals for the people it invited, me included. The day’s sessions were led by Philip Lakin, Zapier’s director of AI transformation, and Alex Maxwell, who runs Zapier’s AI Leaders Lab.

Make it the job, not a side project

Lakin did not argue. In that case, he said, leadership had never really aligned behind it.

“Unless you create the time and you show that you actually value it, what you compensate, what you incentivize, and what you create time for, it’s never gonna happen,” Lakin said.

Zapier’s own structure reflects this. People on its AI transformation guild sit in their lines of business. But 30% to 40% of their job is transformation work, and Zapier removes that share of their other duties. The role has three parts: listening to users, building, and turning the mandate into things the team can measure.

Spreading it thin fails too. Telling a whole company that AI is 10% of everyone’s job gives nobody a reason to make time.

“When everyone’s responsible for something, no one’s responsible for it,” Lakin said.

Start with relief, not efficiency

The usual pitch for AI is efficiency: save time, cut cost. The people doing the work read it differently. They either have no time to experiment, or they suspect that mapping their own workflows is the first step to replacing them.

The pitch that works, Lakin told the room, is relief: something that removes the part of the job people hate, or that “makes you look good”. At the end of the day, one attendee said the question they would take home was exactly that. Ask what people hate about their job, not how they could save time.

TNW has reported on both halves of that fear. Meta’s chief technology officer has said AI productivity gains mean more work, not time off. Hyundai workers voted to strike over automation. Younger workers report feeling guilty about using AI at work at all.

Find the champions by watching, not asking

Someone asked how to identify the people worth investing in, when most of the organisation already has more work than it can carry.

Grady Sokaras, Zapier’s head of sales, gave two routes. You hire the person outright, or you bring a group through a basic workshop and watch. Zapier runs a 90-minute session that teaches a small group to build its first workflow. You see who it clicks for. Then you train those people and make the work 30% to 40% of their job, or all of it.

The company has published its own account of the same approach, including hackathons and internal champions. The theory behind it is that people who know how to build spot better problems, even when they never build the production version themselves.

Market the results or lose the time

“Things do not, especially in operations, things do not internally market themselves,” Lakin said. “They have to be marketed.”

That is how a champion keeps the time. Lakin told the room to go back to the managers who control their workload with the figures.

“Does your manager know that you took this number and made it this number? Why would they ever take time away from you?” Lakin said.

Diagnose which layer is broken

Lakin walked the room through a three-layer diagnostic. Start at the bottom: if people in the lines of business are not pushing ideas up to the AI guild, the problem is enablement and skills.

Then the middle. If they push plenty, but nothing gets access to data or moves across departments, the problem is the cross-functional layer. Ticket triage exists in every line of business, and somebody has to turn that into one pattern instead of ten.

Then look up. If both of those work and nothing changes, either the mandate was too vague, or the level below the chief executive is not translating it into anything measurable. VPs, directors and senior managers control the reward structures, and the mandate rarely reaches their own numbers.

‘A million years away’

One attendee said their company was “a million years away” from creating the kind of role that sits between operations and IT. Kevin Forestell of Zapier pushed back. Given how quickly things are moving, he said, consider halving the time you think you need, or at least hold it lightly.

The person who raised it had already made the point. Nine months ago, they said, their own role did not exist.

Surface-level adoption keeps widening the divide, and investors are funding the training meant to close it. The AI skills firm Multiverse raised $70m in May. Someone has to be given the time, measured on what they do with it, and believed when they show the results.